Curated datasets

Less data prep.
More time for your project.

Choose data for the work you have in mind. Our paid editions bring cleaned files, clear documentation and an independent audit together, so you can spend more time on your analysis, training or evaluation.

Editions from $36.80One-time paymentDownload access for 30 days

For small businesses, applied AI teams and people building on their own. No subscription or sales call required.

Forest Service ecologist Sjana Schanning measuring a tree diameter in snow-covered woods.
USDA FIA fieldwork in Wisconsin. USDA Forest Service photo by Ben Nurre · License.

The source data is openly licensed and free from its original publisher. Each card links to it. The price pays for the preparation and audit, with the source license kept intact.

01

Find an edition

38 editions across 5 focus areas. Available editions appear first and can be bought now. The rest are in preparation, with a planned price and no buy button until their independent audit passes.

General AI · text

TruthfulQA

Questions crafted to elicit common human misconceptions, with correct and incorrect reference answers and sources.

In this edition: 817 questions (generation + MC configs)
License
Apache-2.0
Original source
huggingface.co, free and openly licensed from its publisher
Curated edition
Planned price $36.80. On sale once its independent audit has passed.
Customer service · text with slot annotations

SNIPS NLU Benchmark

Voice-assistant intent and slot-filling benchmark (weather, music, restaurant booking, etc.).

In this edition: ~16k queries across 7 intents (2017-06 custom intents) plus 2016 built-in intents and 2018 Braun et al. extension
License
CC0-1.0 (repository LICENSE)
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $36.80. On sale once its independent audit has passed.
Customer service · text utterances with intent + entity labels

HWU64 (NLU-Evaluation-Data)

Virtual-assistant intent/slot dataset across many scenarios (alarm, calendar, music, IoT, email, etc.).

In this edition: 20,653 annotated utterances, 60 scenario and intent pairs in 18 scenarios
License
CC BY 4.0
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $36.80. On sale once its independent audit has passed.
Customer service · text chat with action/workflow annotations

ABCD (Action-Based Conversations Dataset)

Customer-service chats for a fictional retailer where agents must follow policy workflows and take system actions (e.g. verify identity, issue refund).

In this edition: 10K+ dialogues, 55 user intents, train/dev/test
License
MIT (repository LICENSE, Copyright 2021 ASAPP Research)
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $36.80. On sale once its independent audit has passed.
Customer service · text with dialogue acts

STAR (Schema-guided Task-oriented dialogues, Rasa)

WOZ dialogues where wizards follow explicit task schemas (bank, insurance, doctor, apartment, etc.).

In this edition: 6,232 dialogues, 24 tasks
License
MIT (repository LICENSE.txt)
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $36.80. On sale once its independent audit has passed.
Customer service · text with dialogue state / act annotations

MultiWOZ 2.2 / 2.1

The standard multi-domain task-oriented dialogue corpus (tourist info: hotels, restaurants, trains, taxi, attractions).

In this edition: 10,416 dialogues across 7 services (3,263 single-service, 6,736 multi-service)
License
MIT (repository LICENSE, Copyright 2019 Pawel Budzianowski)
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $36.80. On sale once its independent audit has passed.
Customer service · text dialogues + grounding documents

doc2dial (IBM)

Goal-oriented information-seeking dialogues where an agent answers using a public-service help document.

In this edition: 4,500+ dialogues over 450+ documents, 4 domains
License
CC BY 3.0 Unported
Original source
huggingface.co, free and openly licensed from its publisher
Curated edition
Planned price $36.80. On sale once its independent audit has passed.
Customer service · text dialogues + grounding documents

MultiDoc2Dial (IBM)

Goal-oriented information-seeking dialogues where an agent answers using several public-service help documents in one conversation.

In this edition: About 4,800 dialogues (21,000+ training turns)
License
CC BY 3.0 Unported
Original source
huggingface.co, free and openly licensed from its publisher
Curated edition
Planned price $36.80. On sale once its independent audit has passed.
Fishing · text (XML)

eCFR Title 50 - Wildlife and Fisheries

Full text of US fishery and wildlife regulations: Magnuson-Stevens implementing rules, quotas, gear and closed-area rules.

In this edition: Single XML for all of Title 50 (incl. Parts 600-699 fishery regulations)
License
Public domain (US federal regulations, government work)
Original source
www.ecfr.gov, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Fishing · tabular (JSON/CSV via REST)

NOAA AFSC Groundfish Assessment Program bottom trawl survey (GAP_PRODUCTS haul/catch/CPUE)

Standardized Alaska groundfish trawl survey data: haul metadata, catch weights/counts and CPUE by species.

In this edition: 35,561 hauls, 900,444 catch rows, 997 taxa, 1982 to 2026
License
Public domain (US Government work): 'no legal restrictions on access to the data. They reside in public domain and can be freely distributed.'
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Fishing · tabular (REST JSON / CSV export)

NOAA Fisheries One Stop Shop (FOSS) commercial landings

US commercial fisheries landings volume and ex-vessel value by year, state and species.

In this edition: Annual landings by state x species, 1950-present; hundreds of thousands of rows
License
Public domain (US Government work, 17 U.S.C. 105)
Original source
www.fisheries.noaa.gov, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Fishing · tabular (Excel/CSV/R data)

RAM Legacy Stock Assessment Database v4.66

Global compilation of fish and invertebrate stock-assessment time series and reference points from management agencies.

In this edition: 111.9 MB zip; roughly 1,200+ stocks worldwide
License
CC BY 4.0
Original source
www.ramlegacy.org, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Fishing · images + bounding boxes

NOAA Labeled Fishes in the Wild v1.1

Small benchmark of unconstrained underwater images (mostly rockfish) with bounding-box annotations.

In this edition: 423 MB zip; ~1,000+ labeled fish images plus negatives and test video
License
US Government work, public domain (17 U.S.C. 105); citation requested
Original source
www.fisheries.noaa.gov, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Forestry · tabular

BAAD - Biomass And Allometry Database for woody plants

Global database of woody plant biomass, size and allometry (height, diameter, leaf/stem/root mass).

In this edition: 20,775 plant records, 223,924 measured values, 678 species, 176 studies
License
CC0-1.0
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Forestry · tabular

Canada National Forestry Database (NFD)

National statistics on wood supply and harvest volumes, forest fires, forest products, silviculture, revenues, insect defoliation and pest control by jurisdiction.

In this edition: 12.5 MB (v3.0.0), multi-decade tables
License
Open Government Licence - Canada
Original source
open.canada.ca, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Forestry · tabular / geospatial points

FPA FOD - Spatial wildfire occurrence data for the US, 1992-2020 (6th ed.)

Wildfire occurrence database with discovery date, final size, cause and location for US wildfires 1992-2020.

In this edition: 2.3 million geo-referenced wildfire records (180 million acres burned)
License
US Government data, no restrictions: 'can be used without additional permissions or fees' (citation requested)
Original source
www.fs.usda.gov, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Customer service · text with slot annotations

Taskmaster 1, 2 and 3

Google's three Taskmaster task-oriented dialogue releases: spoken and written dialogs across flights, food ordering, hotels, movies, music, restaurants, sports and movie ticketing.

In this edition: 54,293 dialogs across three releases
License
CC BY 4.0
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Customer service · speech + transcripts + intent labels

MInDS-14 (PolyAI)

Multilingual spoken intent detection dataset in the e-banking customer-service domain.

In this edition: 8,168 recordings (24.6 hours of audio), 14 language varieties, 14 intents; 0.58 GB
License
CC BY 4.0
Original source
huggingface.co, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Forestry · LiDAR point clouds (UAV ALS) with instance + semantic labels

FOR-instance

Benchmark for semantic and instance segmentation of individual trees in UAV LiDAR, with dev/test splits.

In this edition: 0.86 GB; 1,130 trees in 32 plots, 5 collections (Norway, Czech Republic, Austria, New Zealand, Australia)
License
CC-BY-4.0
Original source
zenodo.org, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Fishing · tabular

ICES DATRAS trawl survey database (NS-IBTS, BITS, etc.)

Standardized international fishery-independent trawl surveys in the NE Atlantic with catch per haul, length distributions, age data and CPUE indices.

In this edition: Multi-decade (NS-IBTS from 1965) haul-level, length-frequency and age (HH/HL/CA) records across ~20 surveys; millions of rows
License
CC BY 4.0 (for open access data), per ICES Data Policy
Original source
www.ices.dk, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Customer service · tabular records (product, issue, company, state, response type)

CFPB Consumer Complaint Database

Official CFPB database of consumer complaints about financial products and services, with product, issue, company and company response categories.

In this edition: 19,203,798 complaints, December 2011 to November 2026, 15 columns; 0.94 GB
License
US federal government work (public domain in the US); CFPB states the data is 'freely available for anyone to use, analyze, and build on'
Original source
www.consumerfinance.gov, free and openly licensed from its publisher
Curated edition
Planned price $105.60. On sale once its independent audit has passed.
Customer service · speech (two-channel audio) + transcripts + metadata

Harper Valley Bank (Gridspace-Stanford)

Simulated banking call-center conversations (balance checks, bill pay, branch hours) recorded by real speakers.

In this edition: 1,381 conversations, 22.8 hours of conversation audio, 59 speakers; 2.6 GB
License
CC BY 4.0 (repository LICENSE)
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $156.00. On sale once its independent audit has passed.
Forestry · satellite imagery + class labels

ForestNet (deforestation drivers, Indonesia)

Forest loss events in Indonesia labeled by driver (plantation, smallholder agriculture, grassland/shrubland, other).

In this edition: 2,756 forest-loss events
License
CC-BY-4.0
Original source
stanfordmlgroup.github.io, free and openly licensed from its publisher
Curated edition
Planned price $187.20. On sale once its independent audit has passed.
Forestry · aerial RGB imagery, canopy height rasters, LiDAR point clouds, bounding boxes

NeonTreeEvaluation Benchmark (data)

Benchmark for individual tree crown detection across 22 NEON sites in the US, with RGB images, canopy height rasters, LiDAR point clouds and hand-annotated crowns in evaluation tiles.

In this edition: 3.1 GB; 6,656 RGB, canopy height and LiDAR files, 30,975 tree crown boxes
License
CC-BY-4.0
Original source
zenodo.org, free and openly licensed from its publisher
Curated edition
Planned price $187.20. On sale once its independent audit has passed.
Forestry · UAV RGB imagery, bounding boxes, tabular field inventory

ReforesTree

Tropical agroforestry carbon-stock dataset pairing drone imagery with ground-truth tree inventory (diameter, height, species group, biomass).

In this edition: 2.3 GB; 4,663 field-measured trees, 103 drone image tiles, 6 plots
License
CC-BY-4.0
Original source
zenodo.org, free and openly licensed from its publisher
Curated edition
Planned price $187.20. On sale once its independent audit has passed.
Fishing · images with classification/count/point-localization/segmentation labels

DeepFish

Fish-habitat image benchmark for counting, localization and segmentation.

In this edition: 6.8 GB; 38,805 images, 3,127 localization and 620 segmentation annotations
License
CC BY 4.0 (data; GitHub code repo is MIT)
Original source
alzayats.github.io, free and openly licensed from its publisher
Curated edition
Planned price $187.20. On sale once its independent audit has passed.
Fishing · images + bounding boxes (COCO Camera Traps JSON)

NOAA Puget Sound Nearshore Fish 2017-2018

Underwater images of fish and crab species from five mesohabitats at Puget Sound aquaculture sites (Scientific Data 2023).

In this edition: 5.1 GB; 49,695 images, 67,968 annotated boxes
License
CC BY 4.0
Original source
data.mendeley.com, free and openly licensed from its publisher
Curated edition
Planned price $187.20. On sale once its independent audit has passed.
Forestry · geospatial vector polygons

NBAC - National Burned Area Composite (Canada), 1972-2024

Canada's national annual burned-area perimeter time series from the Canadian Forest Service.

In this edition: Annual burned-area polygons for all of Canada, 1972-2024 (shapefile)
License
Open Government Licence - Canada (per NBAC ISO metadata and CWFIS datamart)
Original source
cwfis.cfs.nrcan.gc.ca, free and openly licensed from its publisher
Curated edition
Planned price $187.20. On sale once its independent audit has passed.
Forestry · geospatial vector

National Forest Inventory woodland map GB 2024 (Forestry Commission)

Official map of all forest and woodland in Great Britain with interpreted forest type, updated 2026.

In this edition: All GB woodland >= 0.5 ha (national polygon layer)
License
Open Government Licence (UK)
Original source
www.data.gov.uk, free and openly licensed from its publisher
Curated edition
Planned price $187.20. On sale once its independent audit has passed.
Forestry · aerial CIR imagery, Sentinel-1 SAR, Sentinel-2 multispectral, multi-label species tags

TreeSatAI Benchmark Archive

Multi-sensor, multi-label tree species classification benchmark built from official forest inventory stand data.

In this edition: 16.3 GB; ~50k image patches, 20 European tree species classes
License
CC-BY-4.0
Original source
zenodo.org, free and openly licensed from its publisher
Curated edition
Planned price $440.00. On sale once its independent audit has passed.
Forestry · aerial LiDAR point clouds + RGB-NIR imagery

PureForest (IGN France)

Tree species classification benchmark for French monospecific forests pairing very-high-res imagery with dense Lidar per patch.

In this edition: 144 GB; 135,569 patches (50x50 m), 18 species in 13 classes, 449 forests in 40 departments
License
etalab-2.0 (Licence Ouverte / Open Licence 2.0)
Original source
huggingface.co, free and openly licensed from its publisher
Curated edition
Planned price $440.00. On sale once its independent audit has passed.
Forestry · TLS point clouds + quantitative structure models

Wytham Woods TLS individual trees and QSMs

Segmented individual tree point clouds and QSMs from Wytham Woods, used to estimate tree volume and biomass carbon.

In this edition: 42.8 GB
License
CC-BY-4.0
Original source
zenodo.org, free and openly licensed from its publisher
Curated edition
Planned price $440.00. On sale once its independent audit has passed.
Forestry · geospatial vector + stand attributes (species, volume, age, height, proposed operations)

Finnish Forest Centre (Metsakeskus) open forest resource data (metsavarakuviot)

Nationwide stand-level forest resource data for Finnish private forests.

In this edition: ~13.9 million ha of private forests at stand level
License
CC-BY-4.0
Original source
www.metsakeskus.fi, free and openly licensed from its publisher
Curated edition
Planned price $440.00. On sale once its independent audit has passed.
Fishing · sonar video frames + MOT/COCO/YOLO annotations

Caltech Fish Counting (CFC) sonar video dataset

Benchmark for detecting, tracking and counting migrating salmon in imaging sonar.

In this edition: ~132.8 GB; 1,567 clips, 500k+ box annotations
License
MIT (Copyright 2022 Justin Kay)
Original source
github.com, free and openly licensed from its publisher
Curated edition
Planned price $440.00. On sale once its independent audit has passed.
Robotics · RGB images, video, segmentation masks, object IDs, defect labels

ARMBench segmentation (Amazon warehouse robotic manipulation)

Amazon's object-centric benchmark from real warehouse robot picks: instance segmentation in clutter, object identification, defect detection, stow success.

In this edition: Segmentation subset: 53k images, 19.8 GB
License
CC-BY-4.0
Original source
armbench.com, free and openly licensed from its publisher
Curated edition
Planned price $440.00. On sale once its independent audit has passed.
Robotics · multi-view RGB(-D), robot state/actions, language task labels

RoboSet (real-world, RoboHive)

Meta/CMU/UW real-world kitchen-scene manipulation dataset (~250 demos per task) released with RoboHive.

In this edition: 28,500 trajectories
License
MIT
Original source
robopen.github.io, free and openly licensed from its publisher
Curated edition
Planned price $440.00. On sale once its independent audit has passed.
General AI · text

Common Pile Project Gutenberg

English public-domain books from Project Gutenberg plus PG19, with headers/footers removed.

In this edition: 71,810 documents, 26.2 GB text
License
Public Domain (per-document metadata)
Original source
huggingface.co, free and openly licensed from its publisher
Curated edition
Planned price $440.00. On sale once its independent audit has passed.
Fishing · images + instance segmentation + length labels

AutoFish

Fine-grained fish segmentation and length-estimation benchmark mimicking catch monitoring (2025).

In this edition: 15.8 GB; 1,496 labeled images, 454 fish, 18,122 annotations, plus 497 calibration images
License
CC BY 4.0
Original source
huggingface.co, free and openly licensed from its publisher
Curated edition
Planned price $440.00. On sale once its independent audit has passed.
02

What the price pays for

The preparation is included. Each edition brings the files, documentation and record of our work into one package. You can see what changed, what stayed and what to check before using it. Here is what an edition contains when it goes on sale.

Files you can inspect and work with

The files are validated. Parquet or JSONL copies are added in a consistent layout, and the README documents the schema, the counts and the known limits of the data.

Duplicates removed and logged

Exact and near duplicates are removed, and each removal is logged. The cleaning report gives the count, including when the count is zero. Records that the source repeats on purpose, for example the same text under two citations, are kept, flagged and held together in one split.

Personal identifiers redacted where present

Contact details of private individuals and of individual staff members found in the data are redacted and counted by type. Some things are kept on purpose and we say so plainly: official shared mailboxes, agency phone lines and addresses that are part of the public record stay, audio editions contain the speakers' voices, and names that are part of the required attribution or of the public record may remain. Each README states what was removed and what was kept. No automated redaction is perfect, so review the data against your own obligations.

Grouped splits, checked for leaks

Training, validation and test splits are grouped by a key that the README names, for example the subpart of a regulation. The audit checks that no group and no identical record sits on more than one side. Other kinds of relatedness can remain, such as records from the same place in different years, and the README names the ones we know of. A lookup table may be left unsplit, and an evaluation-only benchmark keeps its official split.

License and attribution packaged

The original license text, the attribution the publisher asks for, and the source documents that came with the release travel with the files. That helps with keeping commercial use terms clear, because the terms are in the folder you train from.

Audit, change log and checksums

An edition goes on sale only after a separate, independent audit of the files, the license and the points above has passed. The audit record, a change log of every change made to the source release, and a checksum for every data file are included.

One flat price, paid once

Each edition has one price, shown on its card. Pay once through Stripe and download right away, for 30 days. No subscription and no sales call.

What the audit record looks like in practice. In the eCFR Title 50 edition, audited on November 16, 2026, 3,400 source records went in and 3,255 came out: the 145 set aside were empty placeholders with a heading and no text, and they are logged. The report lists 13 redactions (8 e-mail addresses, 3 personal names, 2 phone numbers) and, just as plainly, what was kept: 158 mentions of 26 official shared mailboxes and 344 mentions of agency phone numbers. No duplicate was removed, and 28 records whose text repeats under different citations were kept, flagged and held in one split. The audit found none of the 3,251 citations on more than one side of the splits. When a cleaning step finds nothing, the report says so.

Curation and audit reduce known problems. They do not promise that an edition is free of errors or fit for a particular use, and the limits of the source data still apply.

03

Raw source versus Krummholz edition

The raw source is a good choice if you have the time to prepare it. Here is what changes when you buy the edition.

TopicRaw sourceKrummholz edition
PriceFree from its publisher.One flat price per edition, paid once.
FilesAs published. Formats and layout differ from source to source.Validated files, with Parquet or JSONL copies added in a consistent layout and a documented schema.
DuplicatesAs published.Exact and near duplicates removed and logged, with the count in the cleaning report. Records the source repeats on purpose are kept and flagged.
Personal identifiersAs published.Contact details of private individuals and individual staff redacted where present and counted by type. Official shared mailboxes, agency phone lines, voices in audio and names needed for attribution remain, and the README says so.
SplitsThe publisher's splits, where it provides any.Training, validation and test splits grouped by a documented key and checked for leaks across that key. Other relatedness can remain, and the README names what we know of. Evaluation-only benchmarks keep their official split.
License and attributionOn the publisher's site, for you to find and keep with your copy.Original license text, attribution text and source documents packaged with the files.
Record of what was doneThe publisher's own documentation.Independent audit record, change log and a checksum for every data file.
Getting itFrom the publisher, by its own download method.Download links right after payment, for 30 days.

Working with your own records?

If your project depends on your business data, tell us what you want to do with it. We can discuss cleaning, labeling and preparing a dataset around the task, with a separate project quote.

Discuss a custom dataset
04

Before you buy

What do I get?

One folder per edition: the cleaned data with Parquet or JSONL copies, a README with the schema, counts and known limits, the original license and attribution text (LICENSE.txt), the source documents that came with the release, the change log (KRUMMHOLZ_CHANGES.md), the cleaning report with its counts (CLEANING_REPORT.json), a checksum for every data file (CHECKSUMS.sha256) and the audit record (VERIFICATION.json). You get download access to all of it for 30 days.

Can I use it commercially?

Each card names the license its publisher chose, and the license text ships with the edition. The licenses in this catalog are US Government works (public domain in the United States), other public domain material and dedications such as CC0, open government licences, Creative Commons Attribution, MIT and Apache 2.0. They allow commercial use under their own conditions, such as giving attribution. Some sources carry further conditions, which LICENSE.txt states: for example not presenting the data as the publisher's official edition, possible third-party material inside a government publication, and requests for citation. The packaged license and attribution text helps with meeting those conditions. Your purchase adds no restriction of its own. This is not legal advice: read the license for your use, above all if you work outside the United States.

Why pay if the source is free?

Every card links to the original publisher, where you can download the source for free. A Krummholz edition includes the preparation and audit: file checks, documented cleaning, appropriate splits, source licenses and a record of the changes. It is an option for teams and individuals who would rather spend that time on their project.

What about refunds?

If you cannot download what you paid for, or an edition does not match what this page says about it, contact info@krummholz.ai or call 207-307-0611 with your receipt. Refund requests are handled one by one. A refunded purchase closes its downloads. Nothing here limits rights provided by applicable law.

How do downloads work?

After you pay on Stripe's hosted checkout, you return to a purchase page that lists every file of the edition with a download link, plus a URL list for wget or curl. There is no account to create. Keep that page's address: it is your order reference, and it reopens your downloads with fresh links for 30 days from payment. If we withdraw an edition during that time, for example to rebuild it, its downloads pause until it is back; contact info@krummholz.ai with your receipt about the files or a refund.

05

Purchase terms

What you are paying for. A one-time charge by Krummholz AI LLC for download access to its curated edition of the named dataset, for 30 days from payment. The source data is openly licensed and available from the credited publisher. The price covers Krummholz's curation and audit work. A purchase does not establish exclusivity over public source records.

The source license still applies. Payment does not add a restriction that conflicts with an upstream license. Training, reuse and redistribution rights granted by the source terms remain intact. Keep required attribution, license notices and modification notices. No publisher endorsement is implied.

What the curation is and is not. Curation and audit reduce known problems; they do not guarantee that an edition is free of errors or fit for a particular use. The limitations of the source data still apply. Validate any dataset before you rely on it in production.

Payment and access. Stripe handles hosted payment. Any sales tax that applies is shown at checkout before you pay. After payment you return to a purchase page with your download links. Keep that page's address: it is your order reference for the 30 days. Links refresh each time you open it. If we withdraw an edition during your 30 days, for example to rebuild it, its downloads pause until it is back; contact info@krummholz.ai with your receipt about the files or a refund.

Problems. For a payment, delivery or access problem, contact info@krummholz.ai or call 207-307-0611 with your receipt. Do not include card details. Nothing here limits rights provided by applicable law.